Isolating circles and blobs as a mask in a binary image

I'm trying to isolate the large blobs/circles as a mask in the attached binary image. To get this image, I did use the imfill function to clean up the specific white areas, but I'm wondering how to isolate the large white areas by either size or solid-ness.

I did watch the Image Segmentation Tutorial and it was helpful as I incorporated pieces into my code, which is below. I utilized the regionprops function to get Area and Solidity information, and I am thinking that the ismember function may be the way to go. Although, I haven't been able to isolate the larger white areas. Any thoughts?

%%
% R = imfill(sliderBW, 'holes'); This is the attached image, defined in the workspace as 'R'

%%
R3 = regionprops(R, 'all'); % Using regionprops to get Area and Solidity
numberOfBlobs = size(R3, 1);

%% 
% Below is filtering the Area; 15000 as a threshold was arbitrary
allBlobAreas = [R3.Area];
allowableAreaIndexes = (allBlobAreas > 15000);
keeperIndexes = find(allowableAreaIndexes);
keeperBlobsImage = ismember(R, keeperIndexes);

%Below is filtering the Solidity; 0.5 as a threshold was arbitrary
allBlobSolidity = [R3.Solidity];
allowableSolIndexes = (allBlobSolidity <0.5);
keeperIndexesSol = find(allowableSolIndexes);
keeperBlobsImageSol = ismember(R, keeperIndexesSol);

%%
maskedImage = R; % Re-defining image 'R' again
maskedImage(~keeperBlobsImage) = 0; % Area
maskedImage(~keeperBlobsImageSol) = 0; % Solidity

[KALYAN ACHARJYA]

Can you explain again?

What you have?

Desired result (You can use paint or any other image drawing tool)?

[Aaron Devanathan]

@KALYAN ACHARJYA Sure, happy to clarify!

The image I had attached to the original post was a binary mask I created after thresholding and using the imfill function to fill in the holes. From that image, the desired result is to isolate the large, circular objects and have that be a mask. I used MSPaint to provide a cartoon example below.

From the code in the original post, I tried the regionprops function in order to get some objects statistics, but even after thresholding the values based on 'Area' and 'Solidity', I wasn't creating the desired mask. Happy to provide more clarification!

[darova]

Once you have Area of each blob - sort area and choose a few biggest

Answer 1

[Image Analyst]

Try this:

% Demo by Image Analyst
clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
imtool close all;  % Close all imtool figures if you have the Image Processing Toolbox.
clear;  % Erase all existing variables. Or clearvars if you want.
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 16;

%===============================================================================
% Read in gray scale image.
folder = pwd;
baseFileName = 'Image_Feb2021.jpg';
% Get the full filename, with path prepended.
fullFileName = fullfile(folder, baseFileName);
if ~exist(fullFileName, 'file')
	% Didn't find it there.  Check the search path for it.
	fullFileName = baseFileName; % No path this time.
	if ~exist(fullFileName, 'file')
		% Still didn't find it.  Alert user.
		errorMessage = sprintf('Error: %s does not exist.', fullFileName);
		uiwait(warndlg(errorMessage));
		return;
	end
end
[rgbImage, map] = imread(fullFileName);
% Get the dimensions of the image.  numberOfColorBands should be = 1.
[rows, columns, numberOfColorBands] = size(rgbImage);
% If it's RGB instead of grayscale, convert it to gray scale.
if numberOfColorBands > 1
	grayImage = rgbImage(:, :, 1); % Take red channel.
else
	grayImage = rgbImage;
end
% Display the original image.
subplot(2, 3, 1);
imshow(grayImage, []);
axis on;
impixelinfo; % Let user mouse around and see gray level.
caption = sprintf('Original Gray Scale Image : %s', baseFileName);
title(caption, 'FontSize', fontSize, 'Interpreter', 'none');
impixelinfo;
% Enlarge figure to full screen.
g = gcf;
g.WindowState = 'maximized';
g.NumberTitle = 'off';
g.Name = 'Histogram Analysis';
drawnow;

%===============================================================================
% Display the histogram
subplot(2, 3, 2);
[pixelCounts, grayLevels] = imhist(grayImage);
bar(grayLevels, pixelCounts);
grid on;
title('Original Gray Scale Image Histogram', 'FontSize', fontSize);
xlabel('Gray Level', 'FontSize', fontSize);
ylabel('Pixel Count', 'FontSize', fontSize);

% Get a binary image
mask = grayImage > 128;
% Get rid of white surround.
mask = imclearborder(mask);
% Display the binary mask image.
subplot(2, 3, 3);
imshow(mask, []);
axis on;
impixelinfo; % Let user mouse around and see gray level.
title('Border-cleared Binary Image', 'FontSize', fontSize, 'Interpreter', 'none');
impixelinfo;

% Ask user for one integer number.
defaultValue = 2000;
titleBar = 'Min Blob Size';
userPrompt = 'Enter the minimum size for blobs you would like to extract : ';
dialogBoxWidth = 80;
caUserInput = inputdlg(userPrompt, titleBar, [1, dialogBoxWidth], {num2str(defaultValue)});
if isempty(caUserInput),return,end % Bail out if they clicked Cancel.
% Round to nearest integer in case they entered a floating point number.
minBlobSize = round(str2double(cell2mat(caUserInput)));
% Check for a valid integer.
if isnan(minBlobSize)
	% They didn't enter a number.
	% They clicked Cancel, or entered a character, symbols, or something else not allowed.
	minBlobSize = defaultValue;
	message = sprintf('I said it had to be an integer.\nTry replacing the user.\nI will use %d and continue.', minBlobSize);
	uiwait(warndlg(message));
end

% Get rid of blobs less than that many pixels in area.
mask = bwareaopen(mask, minBlobSize);
% Display the binary mask image.
subplot(2, 3, 4);
imshow(mask, []);
axis on;
impixelinfo; % Let user mouse around and see gray level.
caption = sprintf('With Only Blobs %d or Larger', minBlobSize);
title(caption, 'FontSize', fontSize, 'Interpreter', 'none');
impixelinfo;

% Find out the areas
props = regionprops(mask, 'Area');
allAreas = sort([props.Area], 'ascend')
% Show distribution of areas:
subplot(2, 3, 5:6);
histogram(allAreas);
counts = histcounts(allAreas);
caption = sprintf('Area Distribution of %d Blobs %d Pixels or Larger', length(allAreas), minBlobSize);
title(caption, 'FontSize', fontSize);
xlabel('Area in Pixels', 'FontSize', fontSize);
ylabel('Blob Count', 'FontSize', fontSize);
grid on;

[Aaron Devanathan]

@Image Analyst This was wonderful, thanks! Let me ask you: is there a way to isolate based on shape vs minimum blob size? Perhaps by circularity?

(Edited to fix spelling of "size")

[Image Analyst]

Yes, you can do shape based filtering by inspecting properties like the circularity and solidity

labeledImage = bwlabel(mask);
props = regionprops(mask, 'Area', 'Perimeter', 'Solidity');
allAreas = [props.Area];
allPerimeters = [props.Perimeter];
circularities = (allPerimeters .^ 2) ./ (4 * pi * allAreas);
allSolidities = [props.Solidity];
% Define thresholds
circularitythreshold = 2; % whatever works.
solidityThreshold = .5; % whatever works.
goodIndexes = find((circularities < circularitythreshold) & (allSolidities > solidityThreshold))
% Extract out only the good ones.
newMask = ismember(labeledImage, goodIndexes);
imshow(newMask)

It's untested, so let me know if that works.

[Aaron Devanathan]

@Image Analyst This worked great! The only thing I altered was reversing the less than sign to a greater than sign for the circularities in the line defining the variable 'goodIndexes'. Thank you for your help!

[Image Analyst]

The circularity by my formula is 1 for a circle and grows from there as the perimeter gets more tortuous. For example an asterisk shape will have a circularity of 6 or 10 or more. So if you want circular shapes, you'll want circularities less than about 2 or 3.

If you say circularities > circularitythreshold then you are specifying the least circular, most tortuous shapes.